CImage reverse dimension cx<->cy !

[solvermi]

Only if i load pics taken vertically (so what width is < of height).
Example: Pic “strFoto” have width of 1000px and height of 3000px.

CImage img;
img.Load(strFoto);
int cx = img.GetWidth(); // return 3000!!!!
int cy = img.GetHeight();// return 1000!!!

Some advice please
Thanks

[TubularX]

Are you sure that your image is not corrupt or incomplete? Can you share it?

[solvermi]

Hi TubularX,
Yes i m shure.
I did tests with pics from Windows Phone and Samsung S4 but is equal...and I see "right" is on Smartphones and Windows but Cimage return reverse!
However if I open with Paint and save with other name these apparently equal work with CImage.

I enclose a photo.

Thank U.

[solvermi]

As you see the photo is taken vertically, although well here you see horizontal as with CImage, while various viewers on Windows and even the simple Paint the show correctly.

Thanks

[TubularX]

I tried with the file "WP_20160403_15_42_25_Pro.jpg" that you shared.

The result is as expected:
cx: 600
cy: 338

However, the width is not less than the height in the image you enclosed!

Of course, based on the motive, a human might argue that the width and height are reversed.

But to the computer, the width is always just a range of pixels in the x-axis and the height in y-axis.

[TubularX]

even the simple Paint the show correctly.

Paint also shows it as 600 x 338...

[solvermi]

Hold on...the original photo has
cx: 1728
cy: 3072

As you can see below

[solvermi]

I think the site that compresses

[TubularX]

OK, then there's probably some orientation metadata in your original jpg file.

Check this out:
http://stackoverflow.com/questions/5...ation-metadata

Edit: Sorry, I just noticed the examples above are in Java. I guess you need a C++ solution. I'm not sure how to deal with it using CImage, but I found a solution with QT: http://imonad.com/photo/jpeg-exif-orientation/

Edit #2: It's also possible with GDI+: https://msdn.microsoft.com/en-us/lib...(v=vs.85).aspx

[solvermi]

You think so too ... I think anyone doing a test with your smartphone to have my problema.....io I tried with Windows Phone with Samsung and equals.
I read a bit of the link and is pretty much the same thing
... though that absurd thing ☺
thank you
Hello.

[solvermi]

Then ... the problem is just the orientation of the images inserted in the Exif file format
So would process all the Exif, but, looking at Exif (and trying) I found that:
If Big-Endian (Motorola) encoding file (like on Windows Phone) the byte to 0x38 identifies the orientation, while for Little Endian (Intel) (as on Samsung) byte is 0x4F.
Of the 8 possible values of expected orientation, in my case I need only 3:
If that byte==3 means that the image is rotate of 180.
If that byte==6 means that the image is rotate of 90CW.
If that byte==8 means that the image is rotate of 270CW.

The rest in my case is not interesting …then I avoid to process all the Exif, do I hurt???


Code sample:

unsigned char b[0x4F]
CFile f(filename, CFile::modeRead);
f.Read(b, sizeof(b) );

unsigned char ucRotate = 0;

// Is a Exif file format?
if (b[0]==0xFF && b[1]==0xD8 && b[2]==0xFF &&
b[3]==0xE1 && b[6]==’E’ && b[7]==’x’ && b[8]==’i’ && b[9]==’f’)
{
// is a Big Endian?
if (b[0xC]==’M’ && b[0xD]==’M’)
{
ucRotate=b[0x37];
}

// is a Little Endian?
if (b[0xC]==’I’ && b[0xD]==’I’)
{
ucRotate=b[0x4E];
}

}

f.close();
…

Thanks.